Censored C-learning for dynamic treatment regime in colorectal cancer study
Zishu Zhan, Zhishuai Liu, Cunjie Lin, Danhui Yi, Jian Liu, Yufei Yang
The Annals of Applied Statistics · 2025 · Volume 19 · Issue 2
Abstract
Dynamic treatment regimes (DTRs) represent sequential decision rules for multiple intervention stages. Each rule maps patients’ covariates to optional treatments. The optimal dynamic treatment regime is the one that maximizes the mean outcome of interest if followed by the overall population. Motivated by a clinical study on the treatment of advanced colorectal cancer with traditional Chinese medicine, we propose a censored C-learning (CC-learning) method to estimate the DTR with multiple treatments based on survival data. To address the challenges of multiple stages with right censoring, we modified the backward recursion algorithm to adapt to the flexible number and timing of treatments. We propose a framework for multiple treatments that transforms the optimization problem of multiple treatment comparisons into an example-dependent, cost-sensitive classification problem. With data space expansion and classification techniques, the CC-learning method can produce an interpretable optimal DTR. We theoretically prove the method’s optimality and assess its performance with finite sample simulations. Using our method, we identify the interpretable tree treatment regimes at each stage for the advanced colorectal cancer treatment data from Xiyuan Hospital.
Keywords
classification perspective, dynamic treatment regime, multiple treatments, survival data
Enable JavaScript for citation tools and additional paper interactions.